AI Tooling Cost Management
Context Scoping
Summary
AI session context should be proportionate to the scope of the task.
Standards
- AI tool requests SHOULD contain only context relevant and reasonably necessary to complete the task.
std-ai-context-scoping-01 - Broad, whole-repository context SHOULD be reserved for repository-wide analysis or cases where narrower, targeted context has already failed.
std-ai-context-scoping-02
Implements These Principles
Session Management
Summary
AI tool sessions should maintain bounded context relevant to their current task.
Standards
- A new AI tool session SHOULD be started when switching to an unrelated task if the existing session history is no longer relevant.
std-ai-session-management-01 - Developers SHOULD periodically reset or summarise long-running sessions to prevent unbounded context accumulation.
std-ai-session-management-02
Implements These Principles
Model Selection & Reasoning
Summary
Model choice and reasoning effort should be proportionate to the task's complexity.
Standards
- AI tools SHOULD use automatic or cost-aware model selection by default.
std-ai-model-selection-reasoning-01 - Maximum reasoning effort or extended-thinking modes SHOULD be reserved for complex or ambiguous tasks.
std-ai-model-selection-reasoning-02
Implements These Principles
Custom Instructions & Tooling
Summary
AI tooling configuration should reduce repeated processing of custom instructions and tool definitions.
Standards
- Organisation-wide and project-level custom instructions SHOULD contain only context that applies across the requests that receive them.
std-ai-custom-instructions-tooling-01 - Organisation-wide and project-level custom instructions SHOULD remain stable between requests to benefit from provider-side caching and avoid repeated cost.
std-ai-custom-instructions-tooling-02 - Tools, extensions, or integrations not required for the current task SHOULD be disabled because their definitions may be included in every request regardless of use.
std-ai-custom-instructions-tooling-03
Implements These Principles
Cost Monitoring & Pricing
Summary
AI tooling cost or usage should be monitored and attributed to the responsible team, and usage practices are reassessed after material pricing changes.
Standards
- Cost or usage attributable to AI tooling SHOULD be monitored.
std-ai-cost-monitoring-pricing-01 - Cost or usage attributable to AI tooling SHOULD be attributed to the team responsible for incurring it.
std-ai-cost-monitoring-pricing-02 - The team responsible for AI tooling adoption SHOULD reassess usage practices after a material change to a vendor's billing or pricing model.
std-ai-cost-monitoring-pricing-03
Implements These Principles
Cost Optimisation Review
Summary
Teams should periodically measure and evaluate cost-optimisation practices in their own context.
Standards
- Teams SHOULD periodically measure the effect of AI cost-optimisation practices on their AI tool usage and cost.
std-ai-cost-optimisation-review-01 - Teams SHOULD evaluate each cost-optimisation practice in their own context because its effectiveness may vary by model, workload, and codebase.
std-ai-cost-optimisation-review-02